Apple Intelligence: What Developers Should Actually Build For

Apple Intelligence isn't just Siri improvements. It's a ~3 billion parameter on-device foundation model running on Apple Silicon, quantized to 2-bit for efficiency . After building with the Foundation Models framework (WWDC25), here's what actually matters for cross-platform developers.

Native First: Swift Gets the Full Stack

At WWDC25, Apple released the Foundation Models Framework — direct Swift API access to the same on-device LLM powering Apple Intelligence.

import FoundationModels

// Create a session with stateful context
let session = LanguageModelSession()

// Generate structured output with @Generable macro
@Generable struct Itinerary {
    var title: String
    var days: [DayPlan]
    var summary: String
}

let response = try await session.respond(
    to: "Plan 3 days in Paris",
    as: Itinerary.self
)
// Returns typed Swift object — not raw JSON

Key capabilities exclusive to native

1. Guided Generation (@Generable, @Guide) — guaranteed structural correctness via constrained decoding

2. Streaming snapshots — partial Swift objects that populate progressively

3. Tool calling — model executes your Swift functions autonomously

4. Zero app size impact — model is baked into the OS

Cross-Platform Reality: Gaps Are Real

The LinkedIn discourse calling Flutter "RIP" is hyperbolic, but the concern has merit . iOS 18+ features increasingly assume Swift/SwiftUI.

What cross-platform tools can't access (yet)

- Native Siri integration and system-level Writing Tools

- Real-time streaming snapshots from the foundation model

- On-device model loading with hardware-specific optimizations (Neural Engine, GPU ML accelerators)

However, React Native is catching up faster than Flutter . Callstack released @react-native-ai/apple (preview), providing:

import { foundationModels } from '@react-native-ai/apple';
import { z } from 'zod';

// Structured output with Zod schema
const schema = z.object({
  name: z.string(),
  price: z.number(),
  summary: z.string()
});

const result = await foundationModels.generateText(
  [{ role: 'user', content: 'Analyze this product' }],
  { schema }
);

// Streaming support
const stream = foundationModels.generateStream([...]);
for await (const chunk of stream) {
  console.log(chunk.textDelta);
}

Requirements: React Native 0.80+, New Architecture enabled, iOS 26+ . Vercel AI SDK compatibility means existing patterns translate.

What Actually Requires Native (For Now)

According to Apple's WWDC sessions, these features have no cross-platform bridge

Feature	Framework	Cross-platform Status
Writing Tools integration	UIKitView/NSTextView	iOS only
Genmoji rendering	NSAdaptiveImageGlyph	iOS only
Siri App Intents	App Intents framework	iOS only
Translation API	Translation framework	iOS only
Vision gestures	Vision framework (Swift 6)	iOS only

Translation API example (native only)

import Translation

let config = TranslationSession.Configuration()
let response = try await translator.translate("Hello world", 
    from: .en, to: .es, 
    configuration: config)

Strategic Recommendation: Hybrid Architecture

Don't rewrite everything. Do this instead:

Layer 1: Core UI (Cross-platform)

- Maintain React Native/Flutter for 80% of screens

- Use for standard lists, forms, navigation

Layer 2: Intelligence Features (Native Modules)

- Implement Apple Intelligence features as native Swift modules

- Expose via React Native bridge or Flutter platform channel

// React Native Module Example
class AppleIntelligenceModule : NSObject, RCTBridgeModule {
  static func moduleName() -> String! { return "AppleIntelligence" }
  
  @objc func generateText(_ prompt: String, 
                          resolver: @escaping RCTPromiseResolveBlock,
                          rejecter: @escaping RCTPromiseRejectBlock) {
    Task {
      let session = LanguageModelSession()
      let response = try await session.respond(to: prompt)
      resolver(response.content)
    }
  }
}

Layer 3: Progressive Enhancement

- Detect Apple Intelligence availability via FoundationModels.isAvailable

- Show enhanced features only on supported devices (iPhone 15 Pro+, M-series iPads/Macs)

// Availability check
if case .available = FoundationModels.status {
    // Show AI features
} else if case .unavailable(let reason) = FoundationModels.status {
    // Fallback to basic UI
}